 & Static Estimate & Pre Trends test & Obs. \\
 Sample restrictions \\
 1. Removing Paris & 0.0115** & [p=0.426] & 39014 obs \\
 & (0.0051) & & (3101 pairs) \\ 
 2. Removing R&D sector & 0.0126*** & [p=0.655] & 46087 obs \\
 & (0.0047) & & (3641 pairs) \\ 
 3. Removing chemistry and pharma sectors & 0.0133*** & [p=0.645] & 43976 obs \\
 & (0.0050) & & (3459 pairs) \\ 
 Controlling for trends
 1. Pre-reform dep. var. linear trend & 0.0108** & [p=0.772] & 44332 obs \\
 & (0.0045) & & (3177 pairs) \\ 
 2. Pre-reform dep. var. trend 4th deg. poly & 0.0106** & [p=0.735] & 44332 obs \\
 & (0.0044) & & (3177 pairs) \\ 
 3. Pre-reform dep. var. trend 4th deg. poly x shock exposure & 0.0084* & [p=0.374] & 44332 obs \\
 & (0.0045) & & (3177 pairs) \\ 
 4.Flexible discipline trend & 0.0145*** & [p=0.952] & 46958 obs \\
 & (0.0053) & & (3705 pairs) \\ 
 5.Flexible science prox. trend & 0.0121** & [p=0.725] & 46958 obs \\
 & (0.0048) & & (3705 pairs) \\ 
 Alternative inference (p-value) \\
 1. Baseline & [p=0.013] & [p=0.725] \\
 2. Clustered at the CZ level& [p=0.024] & [p=0.715] \\
 3. Clustered at the unit level &[p=0.013] & [p=0.753] \\
 4. Twoway clustering & [p=0.019] & [p=0.681] \\
 5. Randomization based inference & [p=0..006] & [p=0..857] \\
